Geologic carbon sequestration (GCS) has drawn increasing consideration as a promising method to mitigate the adverse impacts of climate change. Most issues related to the safety and security of geological CO2 storage arise from the fact that, at typical temperature and pressure conditions encountered in the terrestrial crust, CO2 is less dense than the resident formation fluids. Accordingly, CO2 will experience an upward buoyancy force in most subsurface environments and will tend to migrate upwards whenever permeable pathways are available. Eventually, leaking CO2 could reach shallow zones with valuable groundwater resources and dissolve there. The concern then is that CO2 dissolution causes increased water acidity, which can mobilize trace contaminants present in natural sediments.[3, 4] Another possible hazard is that of widespread fluid pressure increase arising from the injection process. Because of the large volumes of CO2 that need to be sequestered for industrial-scale GCS, the zone of elevated pressure during and after injection can extend many kilometers from the injection site, considerably farther than the extent of free-phase CO2.[5-7] If permeable conduits exist in this zone of elevated pressure, formation fluids (typically low-quality water with high salinity, hereafter referred to as brine) could move upwards along these conduits and then intrude into overlying groundwater resources.[8, 9] This can lead to groundwater quality degradation, because brines may contain high concentrations of heavy metals, NaCl, and other major ions.
To provide potential Class VI permit applicants with more detailed information, the EPA has developed a series of guidance documents, most of which are available today either in final or draft form. A broad variety of topics is covered, for example, site characterization, injection well construction, AoR evaluation and corrective action, and monitoring. The EPA envisions that these documents will be periodically updated to accommodate new information on technologies, tools, and methods. Updates would also allow revising the guidance documents in response to feedback from regulators and permit applicants as they go through the new regulatory process. At the writing of this paper, the first Class VI permit applications have just recently been brought forward for the EPA review (i.e. by Archer Daniels Midland in Decatur, and by the FutureGen consortium in Morgan County, both in Illinois).
Current Area-of-Review Framework
The GS rule requires that the AoR for Class VI injection projects be delineated as part of the permit application based on computational models that (i) account for the physical and chemical properties of all phases of the injected CO2 stream; and (ii) utilize available site characterization, monitoring, and operational data.[12, 16] As mentioned above, the AoR needs to account for the maximum expected extent of free-phase CO2 and the maximum expected extent of fluid-pressure in the injection formation above a critical threshold pressure. Due to buoyancy, leakage of CO2 will occur when free-phase supercritical CO2 reaches permeable pathways through the caprock, because the CO2 is less dense than the native brine. Any such leakage pathways must be investigated and possibly remediated to ensure lasting storage without CO2 leakage. In contrast, brine requires a pressure gradient to move upwards along a leakage pathway, and in an initially hydrostatic system, the presence of saline fluids at the location of a permeable pathway is not sufficient to pose a risk. Thus, leakage of brine is a concern for drinking-water resources only if the pressure increase due to injection is large enough to drive brine from the injection formation to the elevation of the lowermost USDW. The minimum pressure increase at which a sustained flow of brine upward into an overlying drinking water aquifer occurs is referred to as the critical threshold pressure increase, ∆Pcrit. As discussed below, it is fundamentally the conservatism in calculating ∆Pcrit that can lead to very large AoRs.
In practice, permit applicants will develop adequate computational models to predict the CO2 plume extent and the pressure increase in the reservoir over the lifetime of a project. Model results will be analyzed to determine the maximum region of free-phase CO2 as it evolves with time. At the same time, pressure predictions will be inspected to delineate the maximum region where ∆Pcrit is exceeded. The two regions will be overlain and the AoR is then defined as the area where either one or both of these regions exist (Fig. 1). Bandilla et al. conducted a sensitivity analysis to investigate under which conditions the AoR size is defined by the pressure increase or by the extent of the CO2 plume. They found that in about 50% of all cases studied, the size of the AoR was determined by pressure increase, meaning that about half of all permit applications may involve AoRs that are (much) larger than the maximum extent of the CO2 plume. The larger the injection volume, the more likely that reservoir pressure was the defining factor for AoR size (Bandilla KW, 2013, pers. comm.). Also, the difference in size between a pressure-defined and plume-defined AoR increased with increasing injection volume.
Figure 1. Schematic explaining AoR determination in current EPA guidance. (a) shows a typical scenario with the maximum region of free-phase CO2 smaller than the maximum region of pressure increase above ΔPcrit. (b) shows that the AoR encompasses the area where either one or both of these regions exist. Within the AoR, owners or operators of CO2 injection wells are required to identify any potential conduits for fluid movement from the storage reservoir into USDWs (such as the abandoned well and the fault shown above).
Download figure to PowerPoint
The AoR size may become even more extensive when multiple CO2 storage projects are co-located in the same storage formation. For example, Birkholzer and Zhou evaluated regional-scale pressure build-up in Mount Simon sandstone in response to a hypothetical future carbon sequestration scenario involving CO2 storage from 20 separate major carbon emitters totaling injection of 100 Mt/year for a period of 50 years (equivalent to one-third of the total emissions in the area from large stationary sources). Their simulations indicated that multiple-site storage in Mount Simon would result in a very large continuous region with pressure increase above typical ∆Pcrit values, suggesting that the individual AoRs of neighboring projects can merge to form a very large region of several 10 000 km2. Bandilla et al. used a vertically integrated approach to simulate the AoR for CO2 sequestration into the Mount Simon sandstone. The model included injection of approximately 200 Mt/year at 100 sites for a period of 50 years. Results showed that the AoR would encompass about 200 000 km2 with overlapping AoRs for individual injection operations. It is the very large volumes of CO2 that would have to be stored underground for GCS to be an effective climate mitigation measure that makes us propose a tiered AoR framework. The main benefit of a reduced regulatory burden related to AoR compliance is not to be expected for relatively modest injection volumes, such as those in ongoing CO2 storage operations or those envisioned in pending permit applications, but rather in a future world where GCS is a fully employed method used to mitigate the adverse impacts of climate change with a large number of industrial-scale carbon sequestration projects.
Once the AoR size has been determined for a permit application, the question is no longer relevant in the EPA's guidance documents as to whether a location is included in the AoR because of CO2 leakage concerns or brine leakage concerns, or both. This is important to remember in the discussion about the burden of regulatory compliance because the AoR is directly linked to requirements regarding site characterization, corrective action, and monitoring. For example, the EPA's guidance on site characterization requires ‘compiling as much site characterization as is available on the area delineated within the AoR’. The guidance document on monitoring requires that a testing and monitoring plan be developed ‘considering the boundaries of the AoR’. More specifically on groundwater quality and geochemical monitoring, the documents suggests that the ‘AoR is then used in order to design the monitoring system’.
This brings us to the question of how ∆Pcrit should be determined. In the EPA's guidance on AoR delineation, the definition of a critical threshold pressure increase is conservatively based on the concept of brine rising through on open conduit to the USDW (e.g. an unplugged borehole or well). A schematic of this scenario is shown in Fig. 2, where a leakage pathway connects a deep brine formation with an overlying freshwater aquifer. In a typical hydrogeologic setting, the stratigraphic column would be approximately in hydrostatic equilibrium, and, if the formation brine and the fresh water had the same density, any pressure increase in the injection reservoir would generate leakage up the well because the open conduit offers no resistance to upward flow (i.e. infinite permeability in the conduit). However, due to the higher density of the brine, a pressure increase in the injection reservoir does not necessarily generate continuous leakage. For continuous leakage to occur, the actual pressure perturbation needs to exceed a critical minimum value (i.e. the critical threshold pressure increase, ∆Pcrit) such that the increased fluid column weight due to replacing the fluid in the column by denser brine can be overcome. Otherwise, brine leakage would stop before reaching the freshwater aquifer, and, for a constant pressure increase in the injection reservoir, the system would simply attain a new static equilibrium.[9, 18, 19]
Figure 2. Schematic showing upward brine migration inside an unplugged well caused by pressurization from CO2 injection (from Birkholzer et al.)
Download figure to PowerPoint
As suggested by Nicot et al. and later confirmed by Birkholzer et al., the critical threshold pressure increase can be approximated by simple calculations of the weight of the fluid column in the open conduit before and after the entire column has been replaced by formation brine. For hydrostatic conditions, the relevant site parameters defining ∆Pcrit are the density of the brine in the injection formation (mostly a function of salinity, but also of temperature and pressure) and the vertical distance DB from the injection formation to the USDW. To the extent that these parameters vary spatially in the vicinity of a CO2 injection project, the value of ∆Pcrit may also vary spatially. Note that a higher salinity in the injection formation would generally give a larger ∆Pcrit, which in turn would reduce the size of the AoR. That high salinity correlates with a small AoR is an interesting observation given the fact that in the case of leakage the environmental impact of intrusion of higher-salinity brines into a freshwater aquifer would be more severe.
Let us analyze some of the basic assumptions involved in the AoR concept for brine leakage and critical pressure definition. Wells as potential leakage pathways have been discussed in numerous publications on CO2 storage safety.[8, 20-24] In areas where little oil and gas exploration occurred, there are relatively few existing wells with sufficient depth to cause major concern. In other places, more than a century of extensive oil and gas exploration has resulted in a very large number of existing wells, many of which are deep enough to reach the depth of typical CO2 storage formations. If an exploration well is drilled but not developed as a production well, the open borehole is typically plugged with a series of cement plugs before abandonment. For production, a casing is inserted into the borehole, and cement is emplaced along vertical sections of the annular space between the casing and the rock to prevent flow along the outside of the casing. Cased production wells are later abandoned with a series of cement plugs across or above perforations.[23, 24]
The EPA's open-conduit scenario thus corresponds to an extreme leakage case in which typical abandonment methods have either not been implemented or have completely failed. Such scenarios are possible for wells that have been drilled before adequate regulation was in place. However, these historic wells are often relatively shallow, not necessarily penetrating the injection formation. In contrast, complete failure of wells abandoned to modern standards is an unlikely scenario because an open conduit connecting the deep injection formation to the shallow aquifer would only form if several plugs or several cement completions (or a combination of both if the casing corrodes along uncemented well portions) would be completely degraded.
We conclude that the open-conduit scenario describes an ‘extreme’ risk event of low probability and high impact. While it is very unlikely that such a scenario exists close to any actual injection operations, the environmental consequences would be severe, as brine would be moving up the open conduit and into groundwater resources at extremely high leakage rates. The EPA's regulation explicitly accounts for this scenario (a) by defining the AoR such that it encompasses the entire region where pressure increases might drive brine up an open conduit, and (b) by requesting that any wells in this area that could potentially act as open conduits are identified and remediated. We argue that this often very large region should be treated differently in the regulatory guidance; i.e. we propose a risk-based approach in which ‘extreme’ brine leakage scenarios are associated with a tailored set of site characterization, monitoring, and corrective action requirements that is different from other brine leakage scenarios where small to moderate leakage rates should be expected. How this tiered definition of the AoR could be delineated and defined in a risk-based approach will be further discussed later in this paper.
In this context it is worth remembering that many elements of the current AoR approach for brine leakage have been adopted from the well-established regulatory framework for Class I wells in the EPA's Underground Injection Control (UIC) program, which regulates injection of hazardous and non-hazardous fluids into deep rock formations. For Class I wells, the EPA may require that applicants calculate the AoR using a similar critical threshold pressure approach as discussed above for Class VI wells, or the EPA may decide that the AoR shall encompass an area surrounding the injection well based on a fixed minimum radius. The minimum radius is given as a quarter mile for non-hazardous fluids and two miles for hazardous fluids.[18, 26] Over the many years that the UIC program for Class I wells has been in operation, the majority of permits assigned the size of the AoR based on the fixed-radius requirement rather than on a ∆Pcrit calculation. That a small fixed radius was sufficient in many Class I permit applications was likely due to the often relatively small injection volumes, which meant that the reservoir pressure build-up was small in magnitude and spatial extent. In this regard, making regulatory compliance too burdensome due to the conservative open-conduit assumption for ∆Pcrit was not a concern in most Class I permit applications. The difficulty arises when adopting the Class I approach to the more complicated case of CO2 storage, where buoyant fluids are injected at high volumes and where pressure perturbations occur over very large areas.
We should note that one possible approach for minimizing pressure impacts and reduction in AoR size could be to extract native fluids residing in the CO2 storage reservoir so that additional pore space is provided. Pressure management via brine extraction can provide many other benefits, such as increased storage capacity, reduced failure risk, smaller area for monitoring, less interference with other subsurface activities, and manipulation of the CO2 plume.[28-31] On the other hand, brine extraction requires pumping, transportation, possibly treatment, and disposal of substantial volumes of extracted brackish or saline water, all of which can be technically challenging and expensive. Reservoir management via brine extraction can be a valuable strategy in projects where pressure increase is a real concern, but due to the expected cost brine extraction is unlikely to be become a standard component of large-scale CO2 sequestration projects.
Alternative Well Leakage Scenario
The open-borehole case presented above represents an extreme case of leakage potential, while wells that have been properly plugged and cemented are much less of a concern. However, as discussed below, even properly plugged and abandoned wells (hereafter also referred to as P&A wells) could form preferential pathways for upward migration, albeit typically with much smaller effective permeability and thus much lower leakage rates. Possible leakage pathways along an existing well with poor sealing capacity are shown schematically in Fig. 3, including pathways along the rock-cement interface, along the casing-cement interface, and through fractures or cracks within the cement or the surrounding rock. These pathways may occur from imperfections within the cement, from long-term degradation, or as a result of mechanical impacts during drilling.[21, 33] The features shown in Fig. 3 represent much more realistic brine leakage scenarios, with much higher occurrence probability but also much less impact compared to the open conduit scenario. In a risk-based approach, one would assess what the impact of such realistic brine leakage scenarios would be, i.e. if these could potentially lead to endangerment of USDWs.
Figure 3. Potential leakage pathways along an existing well: between cement and casing (paths a and b), through the cement (c), through the casing (d), through fractures (e), and between cement and formation (f) (from Celia et al.
Download figure to PowerPoint
Several research groups have used Darcy-type flow assumptions to quantify leakage rates for P&A wells, describing the leakage pathways by effective permeabilities for the composite well materials. Effective well permeabilities have been characterized on the basis of field or laboratory measurements, and also on the basis of ‘soft’ information on various well characteristics such as well type, depth, age, and regulatory requirements at the time of drilling. Crow et al. reported on results from vertical interference tests (VIT) to analyze the bulk hydraulic properties along an existing well. VITs measure pressure drops along the outside of the well casing over intervals of several meters, which allows calculating the effective permeability of potential pathways along the rock-cement interface, along the casing-cement interface, and through fractures or cracks within the cement or the surrounding rock. The effective permeability of the well barrier system tested in Crow et al. was found to be in the mD (milliDarcy) range, approximately two to three orders of magnitude greater than the typical permeability of undisturbed cement or shale, but one to two orders of magnitude smaller than the typical permeability in a suitable CO2 storage formation. More recent VIT measurements from nine data sets described in Gasda et al. and Duguid et al. found effective permeability estimates of the well barrier system to be highly variable and ranging from approximately 1 mD to over 1000 mD. In a modeling study on CO2 leakage for the Wabamun Lake area in Alberta, Canada, Nordbotten et al. assigned well segment permeabilities randomly based on a log-normal distribution with a mean permeability of 10 mD (in log10) and a variance of one order of magnitude. Celia et al. translated a qualitative scoring system for well leakage probability[23, 24] into quantitative estimates of effective well permeability, albeit with the focus on CO2 leakage (not brine leakage). According to Celia's study, scores representing ‘low’, ‘medium’, and ‘high’ CO2 leakage potential correlate to permeability ranges of 0.0 to 0.02 mD, 0.02 to 0.5 mD, and 0.5 to 8 mD, respectively, and a permeability range of 8–10 000 mD was associated with scores representing ‘extreme’ CO2 leakage potential. However, as will be shown below, even the upper range of this extreme effective permeability range would not necessarily result in significant brine leakage rates into USDWs.
Another important parameter for risk-based assessment of USDW impact is whether brine migrating upward in a leaky well pathway can interact with intervening formations. In a typical geological situation, the deep CO2 storage formation and the lowermost drinking water aquifer would be separated by multiple layers of alternating low- and high-permeability units, many of which would also be filled with saline water. If the vertical migration of brine occurs at or near the cement-rock interface (which for properly plugged and cemented wells is generally the case over the entire or at least a considerable fraction of the leakage pathway), it is very likely that intermediate brine aquifers would divert a significant fraction of the leaking fluid; i.e. they would act as effective thief zones for further upward migration of brine, essentially protecting the overlying USDWs from leakage impacts.
As pointed out in Birkholzer et al., the concept of a critical threshold pressure that must be exceeded to allow continuous brine flow up an open conduit is not applicable to any of the leakage pathways along P&A wells discussed above, where lateral exchange of fluid and dissolved solids with the surrounding materials (cement, formations) may occur. For example, if formation salinity increases with depth, such lateral exchange reduces the brine salinity and, consequently, the weight of the fluid column in the conduit, thereby reducing the formation pressure buildup necessary to induce sustained flow. An AoR evaluation for such cases cannot be conducted with a simple hydrostatic calculation for ∆Pcrit; rather, it needs to be based on modeling of the system dynamics/transient flow of brine (typically with Darcy-type models for porous or fractured medium flow) and an evaluation of the brine leakage rates into a USDW expected for a given pressure increase in the injection reservoir. Instead of comparing injection pressure with ∆Pcrit to decide whether leakage of CO2 or brine could potentially occur in an open conduit, the decision would be based on risk-based approaches with leakage-pathway models that evaluate the magnitude of leakage and consider the potential severity of impact on groundwater resources. As pointed out below, we believe that Darcy-type brine leakage along P&A wells does not generally result in endangerment of USDWs over the entire region currently included in the AoR definition, at least not if the leakage-pathway models used adequately represent the effective permeabilities along the well and ensure that any thief zones along the way are accounted for.
To get a better feeling for the potential leakage impact in a realistic well leakage situation, we may start querying a very simple leakage setup for a P&A well (Fig. 4). We consider a scenario in which injection of a fluid into a deep saline reservoir occurs for 50 years at an annual rate of 1.85 million m3 (∼1 million metric tonnes), generating a radial pressure perturbation from the initially hydrostatic conditions. A leaky well at an arbitrary distance of 2 km from the injection well allows for upward migration of brine through fractures or cracks outside of the well casing, which endangers the water quality in an overlying shallow freshwater aquifer. In most calculation cases, we assume that an intermediate saline aquifer exists communicating with the leaky pathway and acting as a thief zone. All aquifers (shallow freshwater aquifer, intermediate saline aquifer, and the saline storage reservoir) have a thickness of 100 m, and are separated vertically by low-permeability shale layers of 650 m thickness. The separation distance between the injection formation and the freshwater aquifer is 1400 m.
Because the focus here is on the effects of pressure build-up and brine leakage, we conduct simplified leakage calculations using a recently developed analytical solution for single-phase flow in multilayered systems with diffuse and focused brine leakage. The basic consideration is that the brine pressurization and migration processes outside of the CO2 plume region can be reasonably well described by single-phase flow models – without accounting for local two-phase and variable density effects – simply by representing the injection of CO2 as an equivalent volume of saline water.[39, 40] We start with a calculation for an effective well permeability of 1000 mD over the entire length of the leaky well (with a cross-sectional area for flow of about 0.7 m2, which assumes that leakage occurs along the well materials plus a small damage zone surrounding it). According to Celia et al., this well with its effective permeability of 1000 mD represents the upper range of extreme leakage potential. Yet the cumulative brine leakage volume flowing into the USDW over a 50-year period is only about 270 m3 of brine if no intermediate formation is present. With a thief zone, the cumulative leakage volume reduces strongly to values of about 1.3, 0.15, and 0.017 m3 for intermediate aquifer permeabilities of 10, 100, and 1000 mD, respectively. These numbers demonstrate that (a) brine leakage through a single well pathway will be essentially negligible even though its permeability represents an extreme leakage potential according to the categories defined in Celia et al., and (b) the presence of an intermediate aquifer in communication with the leakage pathway will lead to considerable leakage reduction. We should mention that the pressure in the injection formation increases to a maximum of about 2 bars at the leaky well location, which is a typical pressure range for ∆Pcrit as reported in Nicot et al. and Birkholzer et al. In other words, at a location where the open conduit scenario would require inclusion in the AoR, the more realistic leakage case of brine migration along fractures and cracks would have negligible impact on USDWs.
Above findings are confirmed by modeling studies for brine leakage in more complex and realistic CO2 sequestration situations. For example, Cihan et al. investigated the potential far-field pressure build-up and brine leakage for an industrial-scale CO2 storage project, which assumes injection of five million metric tonnes of CO2 annually into a deep saline formation underlying a multilayered aquifer-aquitard system. The deepest of four freshwater aquifers in the multilayer sequence is separated from the injection formation by three aquitards and three saline aquifers. One of the many simulation cases considers the presence of 36 P&A wells in a 5 × 5 km2 area about 20 km south of the injection location. Each of these wells is assumed to be leaky with an effective permeability of 1000 mD. Results of their study show that the overall magnitude of well leakage is very small, despite the fact that CO2 injection induces strong and far-reaching pressure increases in the storage formation (between 10 and 15 bar in the leaky well field). Brine leaves the storage formation at a maximum annual rate of about 10 000 m3 (or about 0.2% of the annual injection volume), but since the leaky flow paths communicate with all intermediate aquifers, the majority of the brine leaving the storage formation recharges into the nearest overlying saline layers (Fig. 8(c) in Cihan et al.). As a result, the maximum rate of brine entering the deepest of the four USDWs is only about 100 m3 per year (less than 0.002% of the annual injection volume), occurring over a 25 km2 area, or about 3 m3 per year per leaky well. Furthermore, well leakage decreases strongly when injection ends and reservoir pressure declines. These leakage rates and total volumes are too small to cause environmental concern for the freshwater aquifers. In contrast, if leakage was to occur in open high-permeability wells and without communication with intervening saline formations, the leakage rates would be drastically increased – roughly 1.8 million m3 per year (or about 21% of the annual injection volume) leaving the storage formation and being redistributed into the four freshwater aquifers (Fig. 8(a) in Cihan et al.). Similarly strong leakage rates were calculated for another extreme case, where a very large conductive fault connects the injection formation with the overlying USDW (Fig. 8(d) in Cihan et al.). These extreme leakage scenarios would be associated with strong pressure perturbations in the freshwater aquifers (up to 3 bar), which should be easily detectable.
In another relevant study, Celia et al. used semi-analytic methods to study large-scale leakage of CO2 and brine along abandoned wells. The model included a succession of permeable and impermeable formations representing a sedimentary basin stratigraphy, and a large number of abandoned wells. The petro-physical parameters and well locations and depths were based on the Wabamun Lake area in Alberta, Canada. While the well locations and depths were known, the effective permeabilities along the wells needed to be estimated. The permeabilities were assigned using a stochastic approach with mean permeabilities based on the leakage potential approach developed by Watson and Bachu and measurements by Crow et al. The wells were divided into segments, where each segment represented the thickness of a stratigraphic layer along the wells, and different permeabilities were randomly assigned for each segment of a well. Monte Carlo simulations showed that the volumetric brine leakage out of the injection formation after 50 years of injection was around 0.01–1% of the volume of the injected CO2. However, brine leakage to the shallowest aquifer was less than 0.0001% of the injected CO2 volume. Celia et al. attributed this leakage reduction to the impact of the intermediate aquifers that help dissipate the pressure increase, as well as to the fact that a single well segment with low permeability will effectively reduce leakage further up the well. The Celia et al. study confirms that while the injection-induced pressure may cause brine leakage along abandoned wells, the brine volume reaching a USDW is often negligible.
Nicot et al. conducted a similar study of CO2 and brine leakage potential for a CO2 sequestration operation in the Cranfield field, in Mississippi, United States. The highest risk for leakage at this site was expected from several P&A wells in the area, for which extensive cement bond logs (CBL) had been conducted. A CBL is used to analyze the integrity of bonding between the cement and the well casing. All wells intersecting the designated CO2 storage formation (i.e., the lower Tuscaloosa Formation) were categorized into segments of ‘100%’ cement (near perfect bonding), ‘good’ cement (some non-connected areas with a poorer bond), and ‘questionable’ cement, and effective permeabilities were assigned to each category based on measured permeabilities for intact and degraded cement (i.e. 10−6 mD for 100% and good cement, 1 mD for questionable cement, and 10 mD for uncemented intervals). Leakage rates for each well were estimated using a semi-analytical solution for well leakage that takes into account vertical leakage with radial flow from the well into adjacent formations, as controlled by local rock properties. Results show a possibility of non-negligible CO2 leakage for wells with poor-quality cement, albeit at very low volumes (less than 0.0002% of annual injection rate). In contrast, the lack of buoyancy renders brine leakage up the wells negligible as overpressure from CO2 injection rapidly dissipates in permeable layers above the injection reservoir. Essentially no brine flux is predicted to occur in any well interval above the Upper Tuscaloosa Formation (i.e. the permeable formation immediately above the primary seal, which itself is overlain by a thick sedimentary sequence with mudstones and sandstone units), even for the highest estimated well permeability in the area.
Revisiting Area of Review for Brine Leakage
We propose that the current EPA guidance on AoR for pressure-driven brine leakage should be revisited and revised. We are concerned about the size of the AoRs when it is based on the hypothetical existence of completely open conduits, for example open boreholes. These AoRs can be particularly large when industrial-scale injection volumes and/or multiple co-located storage projects could generate extensive regions of pressure build-up in a storage reservoir. A large AoR can make GCS projects unnecessarily expensive since, according to the current EPA guidance documents, the region determined as AoR is associated with specific regulatory requirements regarding site characterization, corrective action, and monitoring activities.
We suggest here that there should be different definitions and different requirements for the AoR, depending on whether the concern is about free-phase CO2 or pressure-driven brine migration, and that these definitions and requirements should take into account what type of leakage pathway is to be considered. The driving forces for CO2 leakage are very different from those for brine migration, and the physics of Darcy-type flow along a P&A well (through well cement or fractured or damaged zones around well casings, e.g. Fig. 3) are very different from those in hypothetical open conduits. As a result, the potential leakage rates and the environmental impacts on USDWs differ significantly. Our suggestion is to define a tiered AoR based on computational model results, one specific to the predicted extent of free-phase CO2, the other two specific to the predicted extent of the pressure increase, with a separate set of requirements for site characterization, monitoring, and corrective action. In most practical applications, the three-tier definition will collapse to a two-tier AoR, as discussed below and illustrated in Fig. 5. It is possible that EPA regulators use best judgment in their permit application reviews to evaluate the risks associated for CO2 leakage differently from the risks associated with pressure-induced brine leakage, but the current guidance documents for AoR delineation do not support this practice. Our tiered AoR proposal is an attempt at initiating discussion about and possibly formulating what might become an improved AoR definition.
Tier 1 AoR for extent of free-phase CO2
The Tier 1 AoR encompasses the region surrounding the injection well(s) defined by the maximum expected extent of free-phase CO2. This is the region of highest risk during and after injection of CO2 because of the potential for mobile supercritical carbon dioxide encountering leakage pathways. It is also the area that experiences the highest pressure buildup from injection, which could cause pressure-driven CO2 and brine leakage. It therefore makes sense that this is the region with the highest standards regarding site characterization, monitoring, and corrective action. The EPA's guidance documents on site characterization and monitoring give a broad overview of recommended activities in the AoR, and we view these as realistic guidelines for characterization and monitoring specific to the Tier 1 AoR.
Tier 2 AoR for brine leakage in open conduits
The Tier 2 AoR addresses concerns about major brine leakage pathways in a region beyond the Tier 1 AoR, i.e. outside of the area where free-phase CO2 is present. The Tier 2 AoR considers the specific low-probability and high-impact scenario of leakage in hypothetical open conduits or unplugged wellbores, and the concept of critical threshold pressure currently described in the EPA's guidance document is appropriate in this case. In many applications (but not always), this AoR would be much larger than the other AoRs, but the burden of meeting the regulatory requirements in terms of site characterization and remedial action can be reduced if these are tailored to those pathways with significant potential for brine leakage. Based on the previous sections, it would be sufficient to identify and if necessary plug all conduits in this area that could fall into the specific category of unplugged or improperly plugged wells. This can be done by one or more of a set of generally inexpensive methods,[44, 45] such as surveys of existing information (e.g. historical records of well leakage problems, photos, well records and other data bases, corrective well actions) and remote sensing (e.g. satellite surveys, magnetic surveys, infrared photography, spatial pattern analysis, vegetation age changes). In addition to a focused effort on finding unplugged and improperly plugged wells as major brine leakage pathways, site characterization in the Tier 2 AoR could be fairly basic with emphasis on the regional hydrogeology of the injection reservoir and the confining units, while detailed petrological, mineralogical, geochemical, geomechanical, and geophysical characterization would be mainly conducted in the Tier 1 and Tier 3 AoRs. Furthermore, monitoring systems for groundwater quality within the Tier 2 AoR could be tailored to detecting brine leakage in open conduits or other highly permeable pathways. And finally, in terms of mitigation, corrective action would be limited to those artificial penetrations that fall into the category of open conduits or unplugged wellbores. In summary, we expect the characterization, monitoring, and corrective action plan targeted to this low-probability high-impact scenario of open-conduit leakage to be much less burdensome than a plan that caters to the entire AoR as currently defined in the EPA's guidance.
Tier 3 AoR for Darcy-type brine leakage along P&A wells
We propose this third AoR tier for brine leakage along all existing wells that are not open conduits as defined above; i.e. these wells have been plugged and cemented to standard but may provide leakage pathways due to imperfections within the cement, from long-term degradation, or as a result of mechanical impacts during drilling. These pathways could occur along the rock-cement interface, along the casing-cement interface, or through fractures or cracks within the cement or the surrounding rock. Based on our own simple calculations and those of others cited in Section 3, the cumulative brine leakage volumes are generally very small (if not negligible) so that USDW water quality impacts should only be expected at high differential pressures. This result is in part due to the typically small effective permeabilities, but can also be caused by the presence of intermediate thief zones that can siphon off a major fraction of the migrating brine. In contrast to the concept for the Tier 2 AoR, where the simple hydrostatic calculation for ∆Pcrit determines only whether leakage may occur or not, the Tier 3 AoR delineation needs to be risk-based considering the magnitude of leakage and the severity of impact on USDWs. We propose here to conduct site-specific well leakage simulations similar to those discussed earlier, and to derive the critical reservoir pressure at which the magnitude of brine leakage into a USDW would exceed a certain threshold level based on the no-endangerment requirement (i.e. the level at which water quality impacts in the USDW could be expected plus a safety factor). Pressure predictions for the storage formation would then be inspected to delineate the region where this critical value is exceeded. As demonstrated in the calculation examples given, this risk-based approach does not necessarily involve very complex modeling tools; in many cases, simple analytical or semi-analytical calculations may suffice. As to the regulatory standards associated with the Tier 3 AoR, the site characterization and monitoring requirements would again be tailored to address the leakage characteristics of concern, which in this case is brine migration along P&A wells and/or minor fracture pathways.
Figure 5. Schematic explaining AoR determination in a two-tier AoR framework. (a) shows a typical scenario with the maximum region of free-phase CO2 being smaller than the maximum region of pressure increase above ΔPcrit. (b) shows the two types of AoR: Tier 1 AoR, the projected region of maximum expected extent of free-phase CO2, which should be the region with the highest standards regarding site characterization, monitoring, and corrective action; and Tier 2 AoR, the projected region with pressure increase above critical pressure, in which the site characterization and monitoring should be much less burdensome, tailored to unplugged wells and other major pathways for brine leakage.
Download figure to PowerPoint
In most practical applications, we expect the impact of brine leakage along P&A wells to be very small if not negligible outside of the region where free-phase CO2 exists, which means the size of the Tier 3 AoR would be smaller than the size of the Tier 1 AoR. In other words, one would be concerned about brine leakage along P&A wells and minor fracture pathways in the region that already has the highest standards with regards to site characterization, monitoring, and corrective action because of the potential for buoyant CO2 leakage. If this is the case, the three-tier definition will collapse to a two-tier AoR (Fig. 5). For demonstration, permit applicants should conduct site-specific brine leakage simulations for a few representative locations along the outer limits of the Tier 1 AoR and show that the magnitude of leakage is negligible outside of this region so that no endangerment of USDWs can be expected. In an area with a long history of oil and gas production, this demonstration needs to account for the additive effect of many wells.
Additional AoR Concepts
Most of this paper's discussion so far has been about wells as leakage pathways. This is indicative of the fact that the EPA's AoR framework is mainly focused on ‘artificial penetrations (e.g. abandoned wellbores)’ and less on natural leakage pathways, both in terms of the delineation of the AoR size as well as the guidance on well integrity assessment and corrective action. It appears that the current EPA framework makes the implicit assumptions that the AoR definition as the larger one of two regions – maximum free-phase CO2 extent and maximum region above critical pressure – has sufficient conservatism to encompass not only all relevant artificial penetrations but also all relevant naturally occurring pathways. In other words, though the critical-pressure concept is based on open conduits such as unplugged boreholes or wells, the assumption is made that the resulting AoR also works for natural pathways such as permeable fractures and faults. While one may question why an AoR delineated based on an open well concept would provide a reasonable estimate for the area in which one needs to identify natural pathways, this practice is probably conservative since the open-conduit assumption considers an extreme hypothetical leakage event, with more severe impact than expected from any natural pathways.
One could envision similar arguments when introducing natural pathways into the proposed tiered AoR concept, but in this case it seems appropriate to distinguish between minor and major pathways. The former would be natural fractures in the caprock that could potentially form connected pathways, typically with relatively moderate permeability. We believe that the impact of fluid leakage through such natural fractures can be conservatively represented in the definition for the Tier 1 AoR (regarding CO2 leakage) and Tier 3 AoR (regarding brine leakage). Analogous to the discussion on brine leakage along P&A wells, we would expect the impact of brine leakage along these minor natural fractures to be negligible (in comparison to leaky wells, natural fracture pathways would have lower likelihood to form continuous pathways between injection zone and USDW). Thus we would again expect that the Tier 1 AoR encompasses the Tier 3 AoR. On the other hand, major pathways for fluid leakage could be large conductive faults that connect the injection reservoir with overlying formations, potentially directly with USDWs. If pressure buildup is sufficient, such faults could provide important pathways for brine leakage, similar in impact to the open conduits considered for the Tier 2 AoR. We therefore believe that the site characterization efforts conducted in the Tier 2 AoR should ensure that all potential major pathways in the area are being identified and characterized, which would not only include unplugged or improperly plugged wells but also large faults. These thoughts lead to an expanded tiered AoR definition as follows:
- Tier 1: AoR for extent of free-phase CO2
- Tier 2: AoR for brine leakage in open conduits and in other major pathways for brine
- Tier 3: AoR for Darcy-type brine leakage along P&A wells and in minor fracture pathways for brine
Fluid injections into the subsurface have also been associated with increased potential for seismic events (i.e. small to moderate earthquakes). This is because the reservoir pressure increase can generate damage in the brittle rocks suitable for CO2 sequestration, which in certain situations may express itself as slip along preexisting faults.[47, 48] There is considerable research underway as to the potential magnitude of induced seismic events (e.g. as a function of fault dimension, pressure buildup, in situ stress conditions)[49, 50] and whether such events could lead to the partial loss of sealing integrity of a CO2 storage system. As the scientific community is debating what induced seismicity means to the future of GCS and how the risk of seismic events can be reduced when following best practice standards[51, 52] or employing mitigating measures such as pressure management via brine extraction, the topic raises an additional question for the EPA's GS rule. How can the current AoR framework be adopted to ensure that all preexisting faults are identified that could potentially be reactivated? Maybe ongoing research on induced seismicity will allow definition of a critical pressure related to fault slip that could define an additional area in a multi-tier framework. This would encompass the area in which all faults experiencing a pressure increase above a certain fault reactivation threshold would need to be detected and characterized.